Posted on August 16, 2015

Another thing that needs to be said in connection with the previous post, specifically the idea of not having loads of JS cleverness on the client side, is websockets. I won’t say more than that, because I know very little about them (but I do have the O’Reilly book to put that right). Anyway, just supposing it were possible to make an HTML-over-websocket version that was smooth enough to use? That’s really much more attractive than the byways I’ve been exploring the last several months.

I have gone back to mpv6. I thought I’d start by upgrading it to work with yesod-1.4. Along the way, I made the possibly insane choice to switch from Prelude to ClassyPrelude. This gave me a big headache of problems to solve, but after a long struggle I did get everything to build again. Currently stuck, though, with yesod devel telling me this:

Starting development server...
<command line>: cannot satisfy -package-id main-inplace

Eventually twigged that this was because I’d added a new executable (a test one) to the cabal file, but omitted to mark its stanza with:

if flag(library-only)
   Buildable: False

Anyway, for rather obscure reasons, I really want to be hacking on C at the moment. So I thought I’d return to libmref. I had a second server which was receiving mrefs from my main one, a modified nqmail delivering them, and a mutt built against libmref to retrieve the messages. For a while I was actually reading a couple of mailing lists this way. But I fell out of the habit, and the client side has bitrotted severely in the last 10 months.

On a brighter note, the mrefs themselves are still arriving, courtesy of mpv6 and nqmail. I must ensure that the version of nqmail I’m using is safe and reproducible. (And publish it.)